reproducing-issues
/reproducing-issues
Zamienia niejasne zgłoszenie błędu w kroki odtwarzania z dowodami i wydaje werdykt, bez poprawiania kodu.
- Autor
- arozumenko ↗
- Repozytorium
- sdlc-skills ↗
- Licencja
- MIT
- Ostatnia zmiana repozytorium
- 26.10.2026
Ocena i bezpieczeństwo
- Pliki w paczce
- 2
- Skrypty
- 0
- Komentarze HTML
- 0
- Adresy URL w SKILL.md
- 0
Brak sygnałów. Repozytorium zmienione 26.10.2026, bez odwołań do wycofanych modeli i API.
MIT, otwarta: można używać i zmieniać, z zachowaniem informacji o autorze.
Co robi i kiedy
Zamienia niejasne zgłoszenie błędu w kroki odtwarzania z dowodami i wydaje werdykt, bez poprawiania kodu.
Oryginalny opis (EN)
Use when a bug is unclear or unconfirmed and needs reproduction before RCA or any fix. Turns a vague bug report into precise, repeatable steps with evidence and a CONFIRMED / CANNOT-REPRODUCE / PARTIAL verdict. Reproduction and documentation only — does not fix code.
Jak działa
- Platform & systems
- Methodology — 5 phases
- Output
- You do NOT
- Related skills
Podgląd SKILL.md
---
name: reproducing-issues
description: Use when a bug is unclear or unconfirmed and needs reproduction before RCA or any fix. Turns a vague bug report into precise, repeatable steps with evidence and a CONFIRMED / CANNOT-REPRODUCE / PARTIAL verdict. Reproduction and documentation only — does not fix code.
license: Apache-2.0
metadata:
authors:
- Artem Rozumenko <artem_rozumenko@epam.com>
version: "0.1.0"
---
# Reproducing Issues
Turn a vague bug report into precise, repeatable reproduction steps backed by
evidence, ending in a clear verdict. **Reproduction and documentation only —
you do NOT fix code.**
## Platform & systems
Findings go on the **tracker ticket** — the source of truth — using the
tracker scout recorded in `.agents/profile.md § Project systems` (GitHub
Issues, Jira via `atlassian-content`, GitLab, Azure Boards, Linear). The
`gh issue …` commands below are the **GitHub reference**; translate per
`.agents/workflow.md § Git host`. In a standalone session with no tracker,
report the same content directly to the user.
## Methodology — 5 phases
### 1. Intake
Read the report in full (`gh issue view <N>` or the tracker equivalent).
Assess: clear steps → follow exactly; partial → fill the gaps; none → explore
the feature area; intermittent → expect multiple timed attempts. Note that
reproduction has started on the ticket.
### 2. Environment setup
Identify the target URL/endpoint/page, auth (user role, credentials),
prerequisite data/state, and client requirements. Reproduction must be
repeatable — document the environment.
### 3. Reproduction attempts
- **UI** → drive the browser with the Playwright MCP (manual-qa's
`playwright-testing` skill on a co-install) or the `browser-verify`
skill: navigate, snapshot for refs, follow the reported steps, screenshot
the failure, capture console errors and network requests.
- **API** → reproduce the failing request with `curl` or a small script;
record status code + body.
- **Logic** → a minimal script calling the function with the edge-case input;
print expected vs actual.
- **Intermittent** → run 5–10×, vary timing and data, document the
success/failure rate.
### 4. Root-cause hints (handoff to RCA)
Gather clues for `root-cause-analysis`: exact console errors + stack traces;
failing requests / unexpected status codes; works-vs-fails patterns;
triggering vs safe inputs; timing sensitivity. Post these technical
observations on the ticket.
### 5. Confirmation gate (required)
Assess the result and post a verdict. **Do not let work proceed to RCA unless
CONFIRMED:**
- **CONFIRMED** — reproduced at least once, repeatable (or intermittent rate
documented), failure clearly captured (screenshot / log / response), steps
precise enough for anyone to follow. Reproduction must land in a **fresh
session/context** — never only in the session where the anomaly was first
observed (earlier interactions there may have created the broken state
themselves). If the repro relies on synthetic event dispatch rather than
native input, say so in the verdict — that fidelity caveat travels to RCA.
Post: reproduction rate, method, exact steps, expected vs actual, evidence,
and RCA hints. → ready for RCA.
- **CANNOT REPRODUCE** — post what was tried and the specific information
needed. → not ready.
- **PARTIALLY CONFIRMED** — reproducible but inconsistent; post the rate and
the conditions under which it fails. → RCA may proceed; flag the intermittency.
## Output
Full report on the ticket; a one-line summary to the user/PM, e.g.
`✅ #N CONFIRMED — 4/5 attempts, trigger: <one sentence>. Ready for RCA.`
## You do NOT
Edit source code, create branches/PRs, make architecture decisions, close
tickets, skip the confirmation gate, or proceed to RCA when not confirmed.
## Related skills
- Fixing is not this factory's job: once a bug is CONFIRMED, file it with the
evidence and walk away (`test-automation-implementation`
references/defect-filing.md). Dev-team factories carry the fix workflows
(`root-cause-analysis`, `bugfix-workflow`).
- **`browser-verify`** for UI reproduction.
Pliki w paczce (2)
- README.md 1 KB
- SKILL.md 4 KB
Wersje
| Repozytorium | Wynik | Licencja |
|---|---|---|
| Ta strona · arozumenko/sdlc-skills | 90 | MIT |
| arozumenko/sdlc-skills | 90 | MIT |
Migawka bazy z 28.10.2026 · wersja paczki 4988db68cdc0 (skrót SHA-256 całej paczki)
Zgłoś problem z tym skillem: szkodliwa treść, błędny opis albo prośba autora o usunięcie.
Podobne skille
- 100issue-reproducer
Zamienia niejasne zgłoszenia błędów w precyzyjne, powtarzalne kroki z dowodami, bez poprawiania kodu, tylko odtwarzając i dokumentując problem. - 90reproduce-and-fix-issues
Odtwarza i naprawia zaufane zgłoszenie błędu lub problemu z wydajnością w ramach wskazanej funkcji. - 100create-bug-report
Tworzy uporządkowane zgłoszenie błędu z poprawnymi krokami reprodukcji, gdy trzeba opisać defekt lub założyć issue. - 87bug-reproduction
Zamienia ogólnikowe zgłoszenie błędu w zweryfikowaną minimalną reprodukcję i test regresji, który najpierw pada, z użyciem git bisect do znalezienia commita wprowadzającego błąd. - 100reproduce-bug
Odtwarza i bada błąd na podstawie zgłoszenia na GitHubie, logów, konsoli i kodu, z zrzutami ekranu z Playwright. - 100report-bug
Tworzy ustrukturyzowany dokument zgłoszenia błędu: kroki reprodukcji, zachowanie oczekiwane i rzeczywiste oraz poziom wagi.